Re: The end of TuxOnIce.

Am Dienstag, 1. November 2011 schrieb Nigel Cunningham:
> Hi everyone.

Hi Nigel.

[…]
> I'm therefore writing to let you all know that I'm planning on calling
> it a day. Before I do that, however, I'd like to get the input of
> others. Does my evaluation of things sound fair to you? Can anyone make
> an argument for not stopping development of TuxOnIce?

While I do use in kernel suspend for quite a while now cause I had issues 
with TuxOnIce for a while and now tend to not bother myself with compiling 
own kernels anymore, I still liked TuxOnIce better.

I think in kernel suspend can still not be aborted. It compresses via LZO, 
but it still does not write as fast as I had known from TuxOnIce. In fact 
hibernation a ThinkPad T520 with 8 GB of RAM, one or two KDE 4.6.5 
sessions can take 15 seconds or something like that with the Intel SSD 320 
thats builtin in. I just tested:

I see about 18 seconds for about 784000 pages for hibernating as well as 
resuming. Thats about 175 MiB/s each way. And in kernel-suspend seems to 
report the speed for *compressing and saving* the data as 784000*4/18/1000 
is about 174 MiB/s. That might be okay for writing the image, but with 
compression I think even there it could be a bit faster still. But reading 
the images should still go quite much faster. I had seen reports with 
400-500 MiB/s here on this list for TuxOnIce and that has been with older 
SSDs. And I know the Intel SSD 320 can serve 250 MB/s on sequential reads 
and about 155 MiB/s on sequential writes - direct I/O, io depth 64, 
measured by fio but with low blocksizes of 2-16 KiB/s. With a low 
compression factor of 30% this could theoretical give at least 300 MiB/s 
on reads and 186 MiB/s on writes. And with larger blocksizes a tad more 
speed at least for reading might be possible. Well still 400-500 MiB/s for 
reads seem out of reach as least unless LZO compresses with a better 
ratio. My estimate was low.

As noted by others: It is your free decision on whether to continue 
TuxOnIce.

Did you consider trying again to contribute to upstream kernel?

I´d actually prefer an enhanced in-kernel-suspend to an out of mailine 
tree TuxOnIce. An in tree TuxOnIce would be really nice, but it does not 
seem thats likely to be accepted by Rafael.

I would like to see the following enhancements to in kernel suspend:

1) aborting hibernation by a key press

2) faster reads, maybe by asynchronous operation

3) faster writes, maybe by asynchronous operation

Anyway: Thanks for all your efforts and good luck with your new projects.

Thanks,
-- 
Martin 'Helios' Steigerwald - http://www.Lichtvoll.de
GPG: 03B0 0D6C 0040 0710 4AFA  B82F 991B EAAC A599 84C7
_______________________________________________
TuxOnIce-devel mailing list
[email protected]
http://lists.tuxonice.net/listinfo/tuxonice-devel